Mineral Pressure Manifestation Prediction Method Based On Machine Learning

With the increase of coal mining depth and intensity,frequent occurrences of coal wall slabs,ground collapse and other mineral pressure manifestations have seriously restricted the safe production of coal mines.Therefore,the analysis and prediction of mineral pressure manifestation in the coal mining process are of great significance to guarantee safe and efficient production of coal mines.At the same time,with the development of internet of things technology,distributed optical fiber monitoring technology has become a new method for monitoring overlying rock deformation during coal mining,this thesis takes the physical similarity simulation test of distributed optical fiber monitoring overburden deformation and periodic pressure as the research background,uses the distributed optical fiber sensing system to obtain the optical fiber measurement point frequency shift value as the data source,and establish the predictive model of mining pressure appearance based on machine learning.The main work of the thesis is as follows:(1)Establish the prediction model of mineral pressure manifestation based on MBCT-SR-RF.To solve the problem of the deformation state of different rock layers above the working face has different influence on the appearance of rock pressure,the average degree of change of optical fiber weighted frequency shift is proposed.This thesis uses the multi-step backward cloud transform(MBCT-SR)algorithm to calculate the digital characteristics of the vertical fiber full-measurement point frequency shift data set:expectation Ex,entropy En and hyper entropy He.The average change degree of the fiber weighted frequency shift and the statistical characteristics of the fiber frequency shift data are taken as the characteristic attributes of the learning sample,and the MBCT-SR-RF mineral pressure manifestation prediction model is constructed.Take Root Mean Square Error(RMSE),Mean Absolute Error(MAE)and Mean Absolute Percentage Error(MAPE)as the evaluation indicators of the prediction model to evaluate the performance of models,and then use BP neural network and Support Vector Regression(SVR)as the contrast method.(2)In order to further improve the prediction accuracy,establish the prediction model of mineral pressure manifestation based on CNN-LSTM.Convolutional Neural Network(CNN)is used to extract features from the original fiber frequency shift data,and then Long and Short Time Memory Network(LSTM)is used to learn the nonlinear relationship between input features and mine pressure manifestation,and establish CNN-LSTM mine pressure manifestation prediction.The model also uses RMSE,MAE,and MAPE as the evaluation indicators of the prediction model,and using CNN and LSTM as comparison methods.(3)The simulation results show that the RMSE,MAE,and MAPE of the MBCT-SR-RF prediction model are respectively 6.2817,5.2529 and 2.2335.Compared with the BP neural network and the SVR prediction model,the RMSE is reduced by 4.3086,1.8575,the MAE is reduced by 3.1857,1.0964,and the MAPE is reduced by 1.164,0.4315,which has higher accuracy and robustness.The RMSE,MAE,and MAPE of the CNN-LSTM combined prediction model are 4.8718,3.1658 and 1.7842,respectively.Compared with the CNN and LSTM prediction models,the RMSE is reduced by 0.3249,2.383,the MAE is reduced by 1.8946,3.3587,and the MAPE is reduced by 0.113,0.7704,the prediction accuracy is better than the other two methods.Finally,on this basis,the universality experiment of MBCT-SR-RF prediction model and CNN-LSTM prediction model is carried out,and the performance of these two prediction models is comprehensively analyzed.The simulation results show that the prediction accuracy of the CNN-LSTM combined model is higher and the universality is better.